Simulation-First Manufacturing: How OpenUSD and Omniverse Are Revolutionizing Production

The Shift from Physical Testing to Simulation

For decades, manufacturing relied on a straightforward but costly assumption: the most reliable test environment was the real world. The classic design-build-test cycle required physical prototypes, on-site commissioning, and extensive trial-and-error on factory floors. Today, that assumption is being overturned by a new paradigm—simulation-first manufacturing. High-fidelity simulation now produces synthetic training data precise enough for production-grade artificial intelligence, enabling perception systems, reasoning models, and agentic workflows to operate effectively in live factory environments.

The key enabler of this transformation is OpenUSD (Universal Scene Description), a framework that has emerged as the connective standard for 3D data interchange. Originally developed by Pixar for film production, OpenUSD allows assets to travel reliably between different design, simulation, and AI training pipelines—eliminating the loss of physics properties, geometry, and metadata that plagued earlier workflows. Manufacturers building on OpenUSD are already reporting measurable gains in speed, accuracy, and cost efficiency.

SimReady: The Foundation for Physical AI

As physical AI becomes integral to industrial operations, manufacturers face a foundational challenge: assets don’t move seamlessly between 3D tools. Every time a computer-aided design (CAD) model is transferred to a simulation platform, crucial information such as material properties, joint constraints, or surface friction can be lost—forcing teams to rebuild models from scratch. To address this, NVIDIA introduced SimReady, a content standard built on OpenUSD that defines exactly what a physically accurate 3D asset must contain to work reliably across rendering, simulation, and AI training pipelines.

SimReady ensures that every asset carries all necessary data—geometric, physical, and semantic—so it behaves consistently whether used in a digital twin, a physics simulation, or a synthetic data generator. Complementing this standard is the NVIDIA Omniverse platform, which provides a physics-accurate, photorealistic simulation layer where AI models are trained and validated before deployment. Together, SimReady and Omniverse allow manufacturers to create “digital rehearsal” environments that mirror real-world conditions with high fidelity.

Four Ways Manufacturers Are Putting the Stack to Work

Several industry leaders have already integrated the NVIDIA Physical AI stack into their operations, demonstrating dramatic improvements in cycle times, accuracy, and cost. Below are two standout examples, with more emerging across the sector.

ABB Robotics Closes the Sim-to-Real Gap at 99% Accuracy

ABB Robotics, a global leader in industrial automation, has embedded NVIDIA Omniverse libraries directly into RobotStudio HyperReality, its simulation platform used by over 60,000 engineers worldwide. The platform represents robot stations as USD files that run the same firmware as their physical counterparts. This makes it possible to train robots, test part tolerances, and validate AI models even before a production line exists.

One of the biggest challenges in robotics is the “sim-to-real” gap—the difference between how a robot behaves in simulation versus the real world. ABB addresses this by generating synthetic training variations at scale, covering scenarios such as different lighting conditions or geometric tolerances that would be impractical to reproduce manually. According to Craig McDonnell, managing director of business line industries at ABB Robotics, “We’ve managed to vertically integrate the complete technology stack and optimize it to a point where we’re now achieving 99% accuracy on the simulated version.” The downstream outcomes are striking: up to 50% reduction in product introduction cycles, 80% reduction in commissioning time, and a 30-40% reduction in total equipment lifecycle cost.

JLR Compresses Four Hours of Aerodynamic Simulation to One Minute

Automotive manufacturer Jaguar Land Rover (JLR) applied the same simulation-first principle to vehicle aerodynamics. Traditionally, computational fluid dynamics (CFD) simulations for aerodynamics can take hours to run, slowing down vehicle development. JLR engineers trained neural surrogate models on more than 20,000 wind-tunnel-correlated CFD simulations across their entire vehicle portfolio. These surrogate models can now predict aerodynamic behavior in about one minute, compared to the four hours required for a full CFD run.

The result is a dramatic acceleration of the design iteration loop: engineers can test hundreds of design variants in the time it once took to test one. Moreover, JLR reports that 95% of its aero-thermal workloads now run on NVIDIA GPUs, leveraging the same simulation infrastructure that also supports AI training. This close integration between simulation and AI allows the company to validate vehicle performance earlier in the design process, reducing reliance on physical wind-tunnel testing and cutting development costs.

The Future of Simulation-First Manufacturing

The examples from ABB Robotics and JLR illustrate a broader trend: manufacturing is entering a simulation-first era where digital testing complements—and in many cases replaces—physical prototyping. By adopting OpenUSD-based standards like SimReady and platforms such as NVIDIA Omniverse, manufacturers can create accurate digital twins, generate synthetic training data for AI, and validate complex systems with near-perfect accuracy. As more companies follow this path, the promise of faster, cheaper, and more flexible production will become the new normal.